Datenvisualisierung 4
Multi-part plots and customisations
2024-01-31
Dodged density plots
We can produce density plots mapped along a catgorical variable by using geom_half_violin() from the gghalves package.
We can also add another geom to add more information to the plot. Let’s add a boxplot.
df_eng %>%
ggplot() +
aes(x = age_subject, y = rt_lexdec) +
geom_half_violin(alpha = .8) +
geom_boxplot()
Maybe we want to move the boxplot so that it’s not overtop of the density plots, and so that it’s not quite as wide. We can do this by setting position to position_nudge(), and width to some value smaller than .75, which is the default width.
df_eng %>%
ggplot() +
aes(x = age_subject, y = rt_lexdec) +
geom_half_violin(alpha = .8) +
geom_boxplot(width = .3, # make less wide
position = position_nudge(x=0.2) # nudge 0.2 along the x-axis
)
This is from a family of options that allow us to alter the position of geoms. For example, Abbildung 1 A and B both show the exact same data, but Abbildung 1 B includes position = position_jitter(0.2) to move overlapping points. This way we get a good idea of how many observations there were across reaction times (y-axis).
df_eng |>
ggplot() +
aes(x = age_subject, y = rt_lexdec) +
geom_point() +
labs(title = "geom_point()") +
df_eng |>
ggplot() +
aes(x = age_subject, y = rt_lexdec) +
geom_point(position = position_jitter(0.2),
alpha = 0.2)+
labs(title = "geom_point(position = position_jitter(0.2))") +
plot_annotation(tag_levels = "A")
If we put all of these plots together, we get a Abbildung 2.
fig_no_colour <-
df_eng %>%
ggplot() +
aes(x = age_subject, y = rt_lexdec) +
geom_point(position = position_jitter(0.2),
alpha = 0.2) +
geom_half_violin() +
geom_boxplot(
outlier.shape = NA,
width = .3,
position = position_nudge(x=0.2))
fig_dens_colour <-
df_eng %>%
ggplot(aes(x = age_subject, y = rt_lexdec, )) +
geom_point(
color = "grey",
position = position_jitter(0.2),
alpha = 0.2) +
geom_half_violin(
aes(fill = age_subject)) +
geom_boxplot(
outlier.shape = NA,
aes(color = age_subject),
width = .3,
position = position_nudge(x=0.2)) +
labs(title = "Violin plot",
x = "Age group",
y = "LDT reaction time (ms)",
fill = "Age group") +
scale_color_colorblind() +
scale_fill_colorblind() +
theme_minimal() +
theme(legend.position = "none")
fig_point_colour <-
df_eng %>%
ggplot(aes(x = age_subject, y = rt_lexdec, )) +
geom_point(
aes(color = age_subject),
position = position_jitter(0.2),
alpha = 0.2) +
geom_half_violin() +
geom_boxplot(
outlier.shape = NA,
# aes(color = age_subject),
width = .3,
position = position_nudge(x=0.2)) +
labs(title = "Violin plot",
x = "Age group",
y = "LDT reaction time (ms)",
fill = "Age group") +
scale_color_colorblind() +
scale_fill_colorblind() +
theme_minimal() +
theme(legend.position = "none")
Positioning errorbar plots
In the second report, you produced errorbar plots, but the errorbars were overlapping. Let’s reproduce something similar using the english dataset. First, we’ll use pivot_longer() to lengthen our data, then we’ll create a summary of reaction times for the lexical decision task and naming task per age group.
If we create an errorbar plot of this data, we get
sum_eng |>
ggplot() +
aes(x = age_subject, y = mean, colour = task, shape = task) +
geom_point() +
geom_errorbar(aes(ymin = mean-sd, ymax = mean+sd))
We can add position = position_dodge(0.2) to force the errorbars to not overlap. We’ll also adjust their width so they’re not so wide (any value lower than 0.75).
sum_eng |>
ggplot() +
aes(x = age_subject, y = mean, colour = task, shape = task) +
geom_point() +
geom_errorbar(aes(ymin = mean-sd, ymax = mean+sd),
position = position_dodge(0.2),
width = 0.2)
But now we’ve left the points behind. We need to also dodge the points, so we add position_dodge() to geom_point(), making sure to use the same value as we did with geom_errorbar().
Customisations
What customisations do you see in the plots below?
fig_default <-
sum_eng %>%
ggplot(aes(x = age_subject, y = mean,
colour = task, shape = task)) +
geom_point() +
geom_errorbar(aes(ymin=mean-sd,ymax=mean+sd))
fig_custom <-
sum_eng %>%
mutate(task = fct_recode(task,
"LDT" = "rt_lexdec",
"Naming" = "rt_naming"),
age_subject = fct_recode(age_subject,
"Young" = "young",
"Old" = "old")) |>
ggplot(aes(x = age_subject, y = mean,
colour = task, shape = task)) +
geom_point(position = position_dodge(0.3),
size = 3) +
geom_errorbar(aes(ymin=mean-sd,ymax=mean+sd),
position = position_dodge(0.3),
width = .3) +
geom_line(aes(group = task,
linetype = task),
position = position_dodge(0.3)) +
theme_minimal() +
labs(
title = "Reaction times per group and task",
x = "Age group",
y = "Reaction time (ms)",
colour = "Task",
shape = "Task",
linetype = "Task"
) +
theme(axis.title = element_text(size = 12,
face = "bold"),
plot.title = element_text(size = 14),
legend.title = element_text(face = "bold"))
Firstly, theme_minimal() was added to each plot to customise the general look. There are a variety of custom themes to try, like theme_bw() or theme_classic(). Try them out.
We can also control individual components of theme by adding customisations with theme(). For example we see in Abbildung 6 A the axis titles are bolded. This was achieved by adding theme(axis.title = element_text(face = "bold)), where axis.title = indicates we want to make a change to the axis titles, element_text() indicates it’s their text that we want to change, and face = "bold" indicates we want to make the text bold. The same was done for legend.title = to make the legend title bold.